This paper proposes a modular framework for integrating AI into government R&D settlement auditing while preserving data privacy. The framework implements 12 anomaly detection rules covering split transactions, Benford's Law violations, and regulatory compliance gaps across 20,000+ expenditure records, with a plugin architecture enabling independent rule auditing and regulatory traceability.
Jeongseung Lee (Sun,) studied this question.